AI Paper Atlas

Research map / Architectures and efficiency / RNNs and normalization

Completed Hyperparameter Transfer across Modules, Width, Depth, Batch and Duration

Bruno Kacper Mlodozeniec, Pierre Ablin, Louis Béthune, Dan Busbridge, Michal Klein, Jason Ramapuram, marco cuturi

ICLR 2026 · Architectures and efficiency › RNNs and normalization · approach: Theory of depth and width · 22 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
use alpha=1/2 residual scaling in CompleteP instead of only alpha=1CompleteP · GPT-style LM on RedPajamaHP transfer across depthNo clear effect
remove QK norms (ablation)CompleteP · GPT-style LM on RedPajamaHP transfer across depthNo clear effect
rescale weight decay by 1/sqrt(kappa) with batch size (ablation)AdamW · GPT-style LM on RedPajamaweight decay transfer across batch sizeImproved
scale LR by 1/sqrt(kappa) with token horizonAdamW · GPT-style LM on RedPajamaLR transfer across token horizonImproved
scale token horizon by increasing batch size only (SDE iso-horizon rule)AdamW · GPT-style LM on RedPajamaLR transfer across token horizonImproved
add per-depth multipliers to per-module learning rates (ablation)per-module LR search · GPT-style LM on RedPajamavalidation lossImproved
uncouple per-layer learning rates beyond Kronecker factorisation (ablation)per-module LR search · GPT-style LM on RedPajamavalidation lossNo clear effect
use optimal per-module HPs instead of optimal global HPs when transferring to 1.3B / 26B tokensCompleteP · GPT-style LM on RedPajamaspeed-up to reach equal loss: 27% speed-up at 420x computeImproved
use per-module search from 50M instead of global random search50M transformer · GPT-style LM on RedPajamavalidation NLL: global best 3.34 nats is substantially higherImproved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy